VCAs instead of POTs

When I first tried to create an audio amp as a keen electronics newbie I just couldn't accept that you throw away almost the entirety of the dynamic range and then gain it X fold. Lifting the noise floor with that gain. I mean the prevalence of input potentiometers preceding gain stages. Mostly they seem to be stereo pots as well, introducing oppurtunities for imbalance, cross talk and "track noise".

One of the first first PCBs I got printed was a stereo baxandal pre-amp with -something db and +20db. Active gain control rather than attenuate and boost taking the noise up with you.

VCA I understand to mean "voltage controller amplifier" and I'm assuming an IC style component that will given an input/output and a voltage signal as to how much to gain it, this "active volume control/gain stage" can be encapsulated in IC land.

In the digital world of course sending a binary 8/16bit would be preferred. However I'll take an analogue voltage control if I have to.

Consider if you digitally process a stream and for whatever reason you drop the "gain" by a factor of 2. The other way you can do that is by shortening the bit width by 1. Dropping "half" the dynamic range. A contrived example is a 5 to 1 mix down stage. You could be tempted to mix E=[signalN/5] but you just cut the dynamic range of all inputs by a factor 5.

How do you do a lossless gain stage? I suppose the answer that jumps out at me from analogue mixing desks best practices is... don't. Try to keep the faders and gain knobs as close to 0db as you can get away with. Flat gain structure to the output stage as much as possible.

Sorry if this post seems odd, I spend a lot of time handling audio as a stream or arrays of numbers and lowering everything by a factor only to indiscrimately gain it a fix amount is like doing: int sample = (sample << 8) >> 8; which is literally a loss in precision of 8 bits.

Am I missing something?
 
Several things, I think.

1. Raising a noise contribution doesn't do much harm as long as it either remains far from dominant or remains well below audibility.
If you have a microphone that produces a few millivolts of signal, amplify it 300 times, attenuate it 10 times and then amplify it 10 times again, the level between the attenuator and the last amplifier is still 30 times as high as at the output of the microphone. If the circuitry in that area is designed for reasonably low noise, it will cause almost no loss of the signal-to-noise ratio. RF engineers have a thing called Friis's formula that expresses this mathematically, see https://en.wikipedia.org/wiki/Friis_formulas_for_noise

2. VCAs are typically translinear circuits that internally work with signal voltage levels of a few kT/q, where k is Boltzmann's constant, T the absolute temperature and q the elementary charge. kT/q is just below 26 mV at room temperature, so a typical VCA actually attenuates a line level signal quite a bit and then amplifies it again. With good design, the dynamic range can still be good enough for the analogue mixing consoles of the 1990's, but if you want spectacular dynamic range numbers, plain old potmeters or digitally controlled amplifiers are probably more suitable.

3. If you process a digital audio signal in a way that causes rounding, it is good practice to apply 2 LSB peak-peak triangular probability density function dither. It worsens the RMS value of the requantization error by typically 10 dB*log10(3) ~= 4.77 dB, but ensures that the requantization error always sounds similar to noise, never like distortion.